Silicon nitride chemical vapor deposition from dichlorosilane and ammonia: theoretical study of surface structures and reaction mechanism

doi:10.1016/S0039-6028(01)01050-0

Surface Science

Volume 486, Issue 3, 10 July 2001, Pages 213-225

https://doi.org/10.1016/S0039-6028(01)01050-0 Get rights and content

Abstract

The structure of a Si₃N₄ film and the mechanism of Si₃N₄ film growth along the [0 0 0 1] crystal direction during chemical vapor deposition have been examined using ab initio MP2/6-31G** calculations. The silicon nitride (0 0 0 1) surface and deposited (chemisorbed) species on this surface were described using cluster models. It was found that the dangling bonds of chemically bound Si and N atoms on the bare surface are relaxed to form additional π bonds or SiN surface double bonds. Energies of reaction and activation energies were calculated and the process of Si₃N₄ film growth was analyzed. It was found that the removal of chemically bound hydrogen from the surface is the rate-controlling step of the deposition process.

Introduction

Silicon nitride is a material of great technological importance because of its electronic and optical properties (high dielectric constant and large band gap), mechanical strength and hardness, and exceptional thermal and chemical stability. Therefore, silicon nitride films are widely used in microelectronics, in solar cells and for mechanical and other applications [1], [2], [3], [4], [5], [6], [7], [8]. The properties and quality of these films are determined by their structure and stoichiometry, which, in turn, strongly depend on deposition conditions such as temperature, pressure, and gas-phase composition [2]. Therefore, a comprehensive understanding of the deposition process could facilitate the development of a reliable method for obtaining silicon nitride films with the desired properties. Such an understanding cannot be achieved without knowledge of the surface structure of a growing film and the mechanism of chemical reactions that can proceed at the surface.

The crystal structure of silicon nitride is well documented [9], [10], [11], [12], [13], [14]. Several papers have been published on the nitridation of Si(1 0 0) and Si(1 1 1) surfaces [15], [16], [17], [18], [19], [20], [21], [22], [23], [24]. In particular, it was shown [23], [24] that the nitridation of a Si(1 1 1) surface with NH₃ gives rise to the formation of β-Si₃N₄ with the Si₃N₄(0 0 0 1) surface. Atomic layer selective deposition of Si₃N₄ on Si(1 0 0) was studied in Refs. [25], [26]. Stoichiometric films were deposited on Si(1 0 0) substrates by the atomic layer controlled growth of Si₃N₄ through self-limiting surface reactions between SiCl₄ and NH₃ [26]. The early stages of nitridation of the (2×1) reconstructed Si(1 0 0) surface with NH₃ (NH₃ adsorption and initial decomposition) have also been studied theoretically by DFT B3LYP calculations [27].

Empirical force fields have been developed for silicon nitride and used for predicting its bulk properties such as equilibrium lattice parameters, phonon dynamics, thermodynamic properties, etc. [28], [29], [30], [31], [32], [33]. Recently, large-scale molecular dynamics simulations have been performed to study the structure, dynamics, and mechanical behavior of cluster-assembled Si₃N₄ and the silicon/silicon nitride interface [34], [35], [36], [37]. Several ab initio calculations have been performed to determine the properties and band structure of bulk silicon nitride [38], [39], [40], [41], [42], [43].

Chemical vapor deposition (CVD) from a mixture of dichlorosilane (DCS) and ammonia is one of the commonly used methods for obtaining silicon nitride films [8]. In our recent theoretical studies we have shown that at lower temperatures the one-step bimolecular reaction between DCS and ammonia, which leads to SiN bond formation and HCl elimination, dominates over the two-step decomposition–insertion reaction path [44], [45]. We also reported preliminary results on the surface chemistry of silicon nitride, including a proposed CVD mechanism with estimated kinetic parameters [44], [45]. In this work, we present the results of a detailed ab initio cluster simulation of the silicon nitride surface structures and surface reactions related to silicon nitride CVD from DCS and ammonia.

Section snippets

Computational details

Quantum chemical calculations were performed using the GAMESS [46] and Gaussian 98 [47] program packages. Geometry optimization was performed using the 6-31G** basis set with the Møller–Plesset second-order perturbation theory (MP2).

Atoms in silicon nitride clusters were divided into two groups: “active” (positions optimized) and “inactive” (fixed coordinates). The active group included selected surface atoms (“active sites”) and any chemisorbed groups or gas-phase molecules reacting with the

Surface structure and relaxation

Depending on the process conditions, Si₃N₄ CVD on a silicon surface may produce either amorphous or crystalline films with different degrees of crystallinity [10], [11], [12], [13], [23], [24]. The crystal structures of both α- and β-modifications of Si₃N₄ consist of nearly planar (0 0 0 1) layers (four and two layers per unit cell for α- and β-Si₃N₄, respectively) [9], [10], [11], [12], [13], [14].

In our modeling study, we constructed surface clusters assuming that the growing film corresponds to

Conclusions

The mechanism of Si₃N₄ film growth was examined by ab initio MP2/6-31G** calculations using the cluster approach. The surface of a growing film was considered to be structured like the Si₃N₄(0 0 0 1) crystal plane. It has been found that the dangling bonds on the bare surface are relaxed to form diatomic >SiN– surface groups. The calculated SiN bond length 1.61 Å is considerably shorter than typical lengths of crystalline Si–N bonds (1.74–1.76 Å), and the surface atoms of these diatomic groups

Acknowledgements

The authors would like to thank Dr. Edward Hall and Dr. William Johnson of Motorola for their support and encouragement. This work has been supported by Motorola's Digital DNA laboratories.

References (51)

W.A. Pliskin
Thin Solid Films
(1968)
C.-E. Morosanu
Thin Solid Films
(1980)
C.-E. Morosanu et al.
Mater. Chem.
(1982)
P. Yang et al.
Ceram. Int.
(1995)
S. Ishidzuka et al.
Appl. Surf. Sci.
(1998)
G. Zhai et al.
Thin Solid Films
(2000)
S. Yokoyama et al.
Appl. Surf. Sci.
(1998)
J.W. Klaus et al.
Surf. Sci.
(1998)
M.E. Bachlechner et al.
J. Eur. Ceram. Soc.
(1999)
A.A. Bagatur'yants et al.
Mater. Sci. Semicond. Process.
(2000)

P.D. Davides et al.

J. Appl. Phys.

(1966)

A. Lekhollm

J. Electrochem. Soc.

(1972)

K.H. Jack et al.

Natural Phys. Sci.

(1972)

R.S. Rosler

Solid State Technol.

(1977)

R.N. Katz

Science

(1980)

K. Kato et al.

J. Am. Ceram. Soc.

(1975)

R. Grun

Acta Cryst. B

(1979)

W.Y. Lee et al.

J. Am. Ceram. Soc.

(1992)

C.M. Wang et al.

J. Mater. Sci.

(1996)

H. Toraya

J. Appl. Cryst.

(2000)

F. Bozso et al.

Phys. Rev. Lett.

(1986)

R. Wolkow et al.

Phys. Rev. Lett.

(1988)

Ph. Avouris et al.

Phys. Rev. B

(1989)

L. Kubler et al.

Phys. Rev. B

(1988)

G. Rangelov et al.

Phys. Rev. B

(1991)

Cited by (35)

First-principles study of electron trapping by intrinsic surface states on β-Si<inf>3</inf>N<inf>4</inf> (0001)
2020, Surface Science
Citation Excerpt :
For atomically-clean β-Si3N4 surfaces prepared in UHV [11], the valence band (VB) is seen to shift rigidly by ∼1.4 eV to lower kinetic energy (i.e., higher binding energy (BE)) following a small exposure to O2, which is consistent with the elimination of negatively-charged surface traps. Several theoretical studies of the chemical properties and electronic structure of β-Si3N4 surfaces have been reported [12–21]. Most of these focus on the (0001), which is not the lowest-energy surface for a bulk single crystal [16], and relatively little is known about the electronic structure of other β-Si3N4 surfaces.
Charge trapping by intrinsic surface states on β-Si₃N₄ surfaces is a relatively unexplored topic in the study of the electronic properties of this material. In the present work, ab-initio density functional theory is used to observe the transfer of an electron from a substitutional-oxygen donor impurity to a coordinatively-unsaturated surface site on the (0001) surface. The process occurs spontaneously, with little or no activation energy, and has a profound effect on surface electronic structure that leads to a range of different states distributed throughout the band gap. The results are discussed in relation to experimentally-observed surface charging that depends on O₂ exposure. The related issue of the β-Si₃N₄ band gap, which has been somewhat controversial, is also discussed in light of recent experimental and theoretical results.
Synthesis and optimization of low-pressure chemical vapor deposition-silicon nitride coatings deposited from SiHCl <inf>3</inf> and NH <inf>3</inf>
2019, Thin Solid Films
Citation Excerpt :
These heterogeneous reactions eventually lead to the formation of the SiN bonds constituting the silicon nitride deposit. A simple model based on a Freundlich's adsorption isotherm [30] or ab initio calculations [52] can be used to simulate the growth of Si3N4 according to such a mechanism. The different effective precursors and therefore the different mechanisms involved in the very low pressure deposition result in lower Ea values than the 230 kJ.mol−1 found with P = 2.0 kPa and DR = 8.
Stoichiometric silicon nitride films were deposited by low-pressure chemical vapor deposition from the SiHCl₃-NH₃-H₂-Ar system in a hot wall reactor at pressures ranging from 0.3 to 2 kPa. The films are amorphous for deposition temperatures up to 1000 °C and crystalline, in the α-form, at 1200 °C and above. A method for evaluating the internal stresses based on the curvature of the silicon substrate wafer and the resulting silicon Raman peak shift was developed. Some amorphous films exhibit high internal tensile stresses that can lead to cracking during deposition depending on the mechanism and effective precursors involved. Residual stresses can thus be reduced and cracking avoided by, in descending order of importance, reducing the concentration of reactive gases through dilution, increasing the deposition temperature and decreasing the total pressure. The effects of these parameters on the intrinsic stresses were related to the amount of residual hydrogen successively incorporated and thermally released during the growth of the coating according to the Noskov's model.
Effects of diverse metal adsorptions on the electronic and optical properties of the β-Si<inf>3</inf>N<inf>4</inf> (2 0 0) surface: A first-principles study
2018, Computational Materials Science
Citation Excerpt :
Moreover, as a semiconductor material with a wide band gap of approximately 4.3 eV [6], silicon nitride possesses an extensive scope of application potential in the field of microelectronics and optoelectronic devices. Because of the higher energy and confusion degree of the interface structure accompanied by significant surface effects, interface adsorption has been regarded as an outstanding candidate approach for providing enhanced properties and great applications in many scientific and engineering fields, especially for the accommodation of band gap, based on the success of the introduction of adsorption, as well as the results of many studies on the surface of β-Si3N4 [7,8]. For example, because of the high physical and chemical activity of the surface and surface of silicon nitride, the study of the adsorption of H2O molecules on the (0 0 0 1) surface was performed; based on the results, it was concluded that the oxide and nitride surfaces could potentially be used as catalysts in the photochemical dissociation of H2O [5,9–11], opening a window for a new approach to hydrogen storage via the adsorptions on the surface.
We use first-principles calculations in this contribution to study electronic structures and optical properties, especially absorption and reflectivity spectra, dielectric constant and loss function of Li, Na, K, Be, Mg, Ca and Al adsorption behaviors on (2 0 0) surface of hexagonal silicon nitride (β-Si₃N₄). The lower adsorption energy of −3.931 eV for Ca-adsorbed surface indicates that it has more excellent structural stability. The seven kinds of adsorptions are all featured by chemisorption with commonly negative value of adsorption energy, implying relatively abated capacity of defending the chemical corrosion formed by metals mentioned above. Band gaps of surfaces with different adsorptions are 0.113 eV, 0.099 eV, 0.113 eV, 0.147 eV, 0.154 eV, 0.151 eV and 0.111 eV for Li-, Na-, K-, Be-, Mg-, Ca- and Al-adsorbed surface structure, respectively, along with 0.143 eV for clean surface of β-Si₃N₄, which leads more efficient measures of potential for obtaining outstanding semiconductor properties for materials. Moreover, absorption spectra curve of surface drops to a lower peak value of 5.46 × 10⁴ cm⁻¹, 4.94 × 10⁴ cm⁻¹ and 3.53 × 10⁴ cm⁻¹ for Na-, Mg- and Ca-adsorption, decreasing by 10.0%, 18.6% and 41.8% in contrast to clean surface, 6.07 × 10⁴ cm⁻¹, respectively, in which the surface carries decreased reflectivity spectra and dielectric loss that are greatly valued and expected in solar cell industry, indicating its broader application potential in photoelectric and microelectronics devices fields.
Structural, electronic and energetic properties of water adsorbed on β-Si<inf>3</inf>N<inf>4</inf> (0 0 0 1) surface: First-principles calculations
2010, Surface Science
Structural, energetic and electronic properties of water molecules adsorbed on β-Si₃N₄ (0 0 0 1) surface, at various coverages, are investigated using density functional theory. At low coverages (θ ⩽ 0.5), it is found that all H₂O molecules undergo spontaneous dissociation forming hydroxyl (OH) and imino (NH) groups where the reactive sites are identified, a result shown for the first time using ab initio theory. For higher coverages (θ > 0.5), only partial dissociation takes place where some of the molecules stay intact being bound via H-bond in good agreement with experimental findings. The driving force for the water dissociation has been identified to be dangling bonds on lower coordinated N and Si surface atoms showing that not all surface atoms are reactive corroborating with previous experimental findings.
Chapter 9 Integrated Approach to Dielectric Film Growth Modeling: Growth Mechanisms and Kinetics
2007, Thin Films and Nanostructures
Citation Excerpt :
In the second way, the cluster is treated as a free (gas-phase) supermolecule, so that this kind of the cluster approximation resembles molecular models with the only difference that such a cluster has commonly no experimental analogs. Clusters of this type have been used in modeling film growth processes in the case of such materials as Si, Si3N4, SiGe, Al2O3, and SiO2[22–24, 31–40]. In the case of ionic materials, the cluster simulating a surface site must be embedded in the surrounding matrix.
Theoretical methods currently used and developed for the description of the epitaxial growth of dielectric and semiconductor thin films are considered. A multiscale approach to film growth modeling is formulated. This approach includes ab initio calculations of the main gas-phase and surface reactions, estimation of the rate constants using transition state or RRKM theory, and kinetic modeling using the kinetic Monte Carlo (KMC) method or formal chemical kinetics. General aspects of quantum-chemical modeling are briefly discussed. Methods for calculating microscopic rate constants of gas–surface and surface reactions based on the transition state or RRKM theory are described in detail. The main aspects of atomistic simulation methods, such as molecular dynamics, Monte Carlo, and KMC in both the lattice and the dynamic versions, are discussed. A KMC method is supplemented with a dynamic relaxation procedure (KMC-DR) in the cases when it is necessary to model irregular growth (defect or amorphous films). Specific technical details of KMC and KMC-DR methods are discussed. A formal kinetic approach to the simulation of film growth is also given. As an example of the specific application of the multiscale techniques, the problem of ZrO₂ ALD on an Si(100) substrate is considered. The simulation of ZrO₂ film growth successfully predicts the structure and composition of the growing film, the formation of defects and accumulation of impurities in the growing film, and the ZrO₂/Si(001) interface structure.
Study of oxygen adsorption on β-Si<inf>3</inf>N<inf>4</inf>(0 0 0 1) by the density functional theory
2006, Chemical Physics
The density functional theory (DFT) B3LYP method is used to theoretically investigate the interaction of O₂ with the β-Si₃N₄ surface (0 0 0 1) at 1200 °C. All the calculations have been performed at the 6-31G^∗ basis set level using H-saturated cluster. From the total energy minimization, the chemisorption on the center of the molecule lying above an Si site and the molecular axis paralleling to the surface is the most stable. After adsorption, the O–O bond is easier to dissociate compared to the free O₂. The electron transferred from the substrate to the O₂ molecule occupies the O₂ anti-bonding orbital, thus leading to a weakening off the bond strength, which is reflected by the elongated O₂ bond length. The changing trend of the O–O population and vibrational frequency is consistent with the change of the O–O bond length. The significant chemisorption energy and the short adsorption bond length indicate that the oxidation occurs on the β-Si₃N₄(0 0 0 1) surface at 1200 °C more easily.

View all citing articles on Scopus

View full text

Silicon nitride chemical vapor deposition from dichlorosilane and ammonia: theoretical study of surface structures and reaction mechanism

Abstract

Introduction

Section snippets

Computational details

Surface structure and relaxation

Conclusions

Acknowledgements

Thin Solid Films

Thin Solid Films

Mater. Chem.

Ceram. Int.

Appl. Surf. Sci.

Thin Solid Films

Appl. Surf. Sci.

Surf. Sci.

J. Eur. Ceram. Soc.

Mater. Sci. Semicond. Process.

J. Appl. Phys.

J. Electrochem. Soc.

Natural Phys. Sci.

Solid State Technol.

Science

J. Am. Ceram. Soc.

Acta Cryst. B

J. Am. Ceram. Soc.

J. Mater. Sci.

J. Appl. Cryst.

Phys. Rev. Lett.

Phys. Rev. Lett.

Phys. Rev. B

Phys. Rev. B

Phys. Rev. B